Real-Time Measurements, Rogue Events and Photon Economics Bahram Jalali jalali@ucla.edu www.photonics.ucla.edu UCLA Science Faculty Research Colloquium May 27, 2009 Also presented at the University of Arizona, Tucson, Az. September 17, 2009, and University of Colorado, Boulder, Co. September 21, 2009 1
Outline Examples of rogue events Their unusual statistics Difficulty in studying them Ultrafast phenomenon provides a good testbed and window Technical challenges and scientific tools needed Time-stretch technology Platform for creating real-time ultrafast t instruments t Extreme phenomenon observed in optics (Nature, December 2007) Taming and harvesting the extreme events Connection with socioeconomics Time-stretch imaging and its application to finding rare cancer cells (Nature, April 30, 2009) 2
Rogue Events, Outliers, Black Swans Singular, rare, but with massive impact A hallmark of complex systems Occur in social, financial and physical science Past behavior does not predict the future Risk analysis models fail October 2008 crash Do ow Jones Ind dustrial Avera age 12000 8000 4000 0 Stock Market Crashes 1992 1996 2000 2004 2008 Year 3
Other Examples of Extreme Events Rogue waves on the high seas Suwa-Maru sinks off the cost of Japan under moderate sea conditions (2008) Believed to be fiction until recently Natural disasters Evolution, survival of the fittestt DNA mutation Cancer: Single rogue cell triggers prostate cancer G.S. Bova et al. Nature Medicine, 2009 Rogue waves Communication Spontaneous noise Wave Height Fractal behavior, cantor set PMD errors in optical communication networks Time Size distribution of innovations 4
Highly Non-Gaussian Probability Distributions Highly Skewed power-law functions Characterized by a Heavy-tail Events much larger than the mean do occur Probability/Risk is much higher than predicted by Gaussian Classification using Mean and S.D. is inadequate Great impact in both finance and engineering Risk / Error Rate models fail Karl Gauss Extreme Value Distribution Probabili ity x exp 2 2 2 Event Magnitude 5
Examples of power laws Phenomena * Phenomenon Approx. Exponent Net worth of Americans 1.1 Website hits 1.4 Book sales 1.5 Probab bility Bill Gates Magnitude Exponent Share of top 1% Share of top 20% 1.1 66% 86% 1.5 22% 58% 2 10% 45% Winner Takes All Survival of the Fittest *Nassim Taleb, Black Swans, 2007 6
It s Difficult to Model and Predict Extreme Events Understanding and modeling is hampered by Scarcity of the events Inability to do experiments Examples: oceanic rogue waves or financial markets Ultrafast phenomena provide an opportunity More data in a short time Laboratory experiments under controlled conditions Enough data to develop models and ability to test them 7
Extreme Value Events in Ultrafast Science Basic ingredients: noise and nonlinearities are present in: Electronic circuits: micro-seconds to nano-seconds Laser optics: pico-seconds to femto-seconds Many similarities il iti between optics and hydrodynamics d Soliton: wave packet that travels without spreading Turbulence occurs in both media Even gravitational field has an equivalent in optics Common mathematics in the form of NLSE Optical power => fluid density Optical Chirp => fluid velocity Soliton on a bed: group stays together Slowest runner Fastest runner 8
Can Ultrafast Phenomena Unveil the Mystery of Extreme Events? Opportunity: Controlled experiments in Table-top testbed Understand the underlying science Learn to engineer and harvest them Challenge: Detecting needle in haystack Requires: fine resolution, non-stop, high throughput Need to understand two classes of instruments: Equivalent time vs. real-time 9
Equivalent-time (sampling) Instruments Workhorse of ultrafast research Slows down the event using strobe light effect Works only for repetitive events Strobe light 2 0-2 Doesn t really slow down the event, just under-samples it Cannot capture singular, transient events Picosolve TM : 600 Gsample/s equivalent time But only 20 Msample/s real-time! 10
Real-time Instruments Fine time resolution, no-stop sampling and long record lengths Holly grail of instrumentation technology What if we can stretch time? Truly slow down events Central theme behind a new class of instruments being developed at UCLA Persistence of Memory, Salvador Dali, 1931 Bahram Jalali & Fredric Coppinger, Data conversion via time manipulation, US Patent # 6,288,659. 11
Our Portfolio of Real-time Instruments Enabled by Time-stretch Time Stretch Digitizer Ali Motafakker Peter DeVore Brandon Buckley Shalabh Gupta Spectrometer Daniel Solli Jason Chou Camera Keisuke Goda Ata Mahjoubfar Ali Motafakker Kevin Tsia 12
Our Approach to Ultrafast Real-time Instrumentation Information to spectrum conversion Spectrum to time conversion Information + Step 1: Femtosecond pulse Group = Velocity Dispersion Map the high speed information onto the spectrum of a femtosecond pulse + = Step 2: Stretched pulse Spectrum is mapped into time Stretch the spectrum into time Digitize the slow waveform (and amplify at the same time) Real-time digitizer Time Stretch allows ultrafast events to be captured in real-time with conventional electronics 13
Time Stretch Enhanced Digitizer Work supported by U. Arizona NSF CIAN ERC & DARPA electrical Fast Electrical signal Time optical Time Stretch electrical Stretched electrical signal Electronic Digitizer Dispersion Chirped Pulse Source Optical modulator PD 45-Gbps data captured using 1.5-GHz digitizer Chirped pulses Stretched pulse F. Coppinger, A. Bhushan, B. Jalali, Electronics Letters, 34 (4), 1998.. Boosts the performance of electronics S. Gupta, A. Motafakker, B. Jalali, Appl. Phys. Lett. 2008 UCLA Photonics Lab 14
Digitizing at 10 Trillion-times per Second in Real-time! 10 Tsample/s Transient Digitizer Sampling Interval = 100 fs 100 fs per conversion step Time stretch extends the performance of electronic digitizers to capture ultrafast events Creates the backend data acquisition needed to capture, store and analyze extreme value events J. Chou, O. Boyraz, D. Solli, and B. Jalali Appl. Phys. Lett. 91, 161105 (2007) 15
Time-stretch is not a Time Lens Time Lens Time Stretch Time-stretch is a Dispersive Analog Optical Link No need to disperse input signal No need to satisfy the lens equation B. Jalali, et al, Nature Photonics 3, 8-10 (2009) 16
Capturing Fast Optical Signals Requirement: fast and non-stop (real-time) Pump-probe Technique Pros: Ultra-fine temporal resolution Long delay Cornerstone of femto-chemistry Cons: Not a real-time technique Only works for repetitive events not applicable to random and rare events Spectrometer Prism CCD Camera Pros: detailed spectral information Cons: Too slow 17
A Brief History of Spectroscopy 1666: Newton uses lens and prism to disperse sunlight onto a screen 1812: Fraunhofer develops diffraction grating 1826: Kirchhoff and Bunsen use gratings to show that each element and compound has its own unique spectrum Newton Fraunhofer Kirchoff & Bunsen Fraunhofer demonstrates spectroscopy Speed ~ KHz Modern Optical Spectrum Analyzer 18
f Time-stretch also Performs Fourier Transform it i. Z t F e e 2. Z total Dispersive Fourier Transformation ~ 2 / 2 dispersion in seconds/hz 2 d f t ~ i 2 Z t 2 2 z F e 2 d Non - zero only if t 2 z frequency to time mapping 2 t z 2 ( stationary phase approximat ion) Real-time Digitizer Time samples D. Solli, J. Chou, B. Jalali, Nature Photonics, 2008 Absorption or Raman Spectrum Group Velocity Dispersion 19 Time 19
Time Stretch Spectrometer Group Velocity Dispersion (GVD) wavelength wavelength Femto-second pulse -470 time 760 Time (ps ) 1990 3220 4450 time Time Single-shot Carbon monoxide 8 pico-meter resolution mission Acetylene 20 km fiber CO Trans OSA 1520 1524 1528 1532 1536 Wavelength ( nm) 0 20 40 60 80 100 Time (ns) P. Kelkar and B. Jalali, Electronics Letters, 1999 J. Chou, D. Solli, B. Jalali, Appl. Phys. Lett, 2008 Spectra acquired at > million times per second Measured without a spectromter Electronic digitizer becomes the spectrometer 20
Outline Needle in the haystack problems and significance of rare and extreme events Difficult to study and mystery surrounding them Ultrafast phenomenon provides a testbed Scientific tools needed for the job and the challenge Real-time vs. equivalent time measurements UCLA s time stretch technology Platform for creating real-time ultrafast instruments Observation of extreme value phenomenon in optics Two examples of recent observations Triggering and harvesting extreme events Ultrafast imaging and the problem of finding rare but diseased cells 21
Observation of Extreme Events in Optics Amplification of laser pulses in silicon Simple but describes key ingredients Generation of supercontinuum pulses in optical fibers Complex phenomenon with connection to oceanic rogue waves 22
Optical Amplification in Silicon First demonstrated at UCLA in 2002 Pump Weak input Silicon Amplified Output 23
Extreme Value Statistics during optical amplification in Silicon Input distribution Output distribution Heavy tail Optical pulses follow the Pareto 80/20 law Photon Economics Unlikely connection between photonics and socioeconomics Both processes governed by the survival of the fittest Due to nonlinear response giving strong preference to the fittest 24 David Borlaug and Bahram Jalali, LEOS 2008. 24
Making Sense of the Extreme Behavior f ( G) G 1 b / 2 2 b 2 2 Exp 2 2 2 I 0 bln( G) 2 Model Measured Heavy tail / power law Heavy tail / power law Results suggest 2 ingredients are necessary 1. Noise / fluctuations in the initial conditions 2. Nonlinear response to the fluctuations 25
Optical Rogue Waves D.R. Solli, C. Ropers, P. Koonath B. Jalali, Nature 2007 26 26
Narrowband input pulse The Experiment Nonlinear fiber Broadband Soliton An nonlinear phenomena Produces white light solitons Resemblance to solitons in water: Common mathematical foundation Both described by the same differential equation Work performed under DARPA PHOBIAC program 27
Optical Rogue Waves D.R. Solli, C. Ropers, P. Koonath, B. Jalali, Nature 2007 28 Rare but intense flashes of white light are produced Simulations can reproduce observed statistics What is causing these rare but dramatic events? 28
Finding the Culprit The culprit is a particular type of noise at the input Intentionally introducing it seeds a rogue pulse Seed is very weak, 1/10,000 of the input Explains unpredictability of rare events 29 D.R. Solli, C. Ropers, B. Jalali, Phys. Rev. Lett., 2008 29
Amplitude and Phase Noise Reduction via Stimulated Supercontinuum Generation Amplitude and phase of the output light become dramatically more stable Phase transition from an disordered to an ordered state of optical radiation D.R. Solli, C. Ropers, B. Jalali, Phys. Rev. Lett., 2008 30
Outline Needle in the haystack problems and significance of rare and extreme events Difficult to study and mystery surrounding them Ultrafast phenomenon provides a testbed Scientific tools needed for the job and the challenge Real-time vs. equivalent time measurements UCLA s time stretch technology Platform for creating real-time ultrafast instruments Examples of extreme value phenomenon observed in optics Understanding, triggering and harvesting the extreme events Ultrafast imaging and the problem of finding rare cells 31
Rogue Cells in Cancer 14 year by John Hopkins Medical School Steven Bova, et al., Nature Medicine, 2009 32
Rogue Events in Biology & Medicine Circulating Tumor Cell (CTC) Breast cancer cells Prof. Dino Di Carlo UCLA Biomedical Engineering Precursors to metastasis Accounts for 90% of cancer mortality Detection of CTC s in blood may offer life saving benefits First commercial device named Top Medical Breakthrough of 2009 Morphologically different than normal cells Potentially detectable via imaging 33
Detecting CTCs: Needle in a Haystack Problem CTCs are extremely rare ~1 cell in milliliter of blood 1 ml = 5 billion cells Need to look at billions of cells! Speed of today s fastest cameras: < 1000 cells per second possible 2 months to find a single cell Need a much faster instrument: camera 34
UCLA s STEAM* Technology World s Fastest Imager K. Goda, K. Tsia, B. Jalali, Nature 2009..\..\..\Pictures\STEAM\STEAM Movies\movie1.wmv * Serial Time Encoded Amplified Microscopy (STEAM) 35
Taking Pictures with an Oscilloscope! 200 150 160ns 163 Single image frame 1 0.8 Time (ns) 70 60 50 40 30 20 10 0 Time Temporal waveform Waveform Captured by Oscilloscope (oscilloscope) 200 100 Voltage (mv) 100 50 Power (mw) 0.6 0.4 0.2 Spectrum Measured by Optical Spectrum Analyzer (Spectrometer) 0-100 Voltage (mv) 0-500 -400-300 -200-100 0 100 200 300 400 Time (ns) 0-200 1580 1585 1590 1595 1600 Wavelength e (nm) K. Goda, K. K. Tsia, and B. Jalali, Nature, 458, 1145 1149 (2009) 36
Real-time Imaging of Micro-fluidic Transport Spatial resolution: ~1 micron (diffraction limited) Frame interval: 163 ns Shutter speed: 450 ps World s fastest continuous camera Image obtained using single pixel detector 37
Time-stamping of Laser Ablation @6Mf Mframes/sec 38
STEAM Imager in the news: 15 minutes of fame 39
Mechanical-scan-free Laser Microsurgery with Real-time Microscopy K. K. Tsia, K. Goda, and B. Jalali, Optics Letters 34 (2009) 40
Laser Microsurgery without mechanical Scanning & with Real-time Imaging Spectral-shower Encoded Confocal Microscopy and Microsurgery (SECOMM) 41 Ablation beam directed to location by tuning the laser wavelength No manual or mechanical motion necessary Micrometer position accuracy Application to Cancer: identification and purging of contaminated (Rogue) cells for Stem Cell Transplant In Collaboration UCLA Medical School 41
Summary Extreme, rogue events appear in diverse range of systems and have dramatic Behavior provides a window into complex systems Studying them has been challenging Ultrafast phenomena provides an opportunity Real-time tools based on the time stretch concept are making this possible Insight gained is offering clues about causes them and means to control harvest them New tools hold promise for studying rare events in biological systems Detection of rogue cells in metastatic cancer 42
Thanks To my Team: Daniel Solli Keisuke Goda David Borlaug Sasan Fathpour Kevin Tsia Shalabh Gupta Ali Motafakker Jason Chou 43